Sources of Toxaphene and Other Organochlorine Pesticides in North

Recently, James and Hites (8), using backward air trajectory analysis, ... built in 1960, located on a fully functioning experimental agriculture farm...
11 downloads 0 Views 767KB Size
Download PDF
Environ. Sci. Technol. 2004, 38, 4187-4194

Sources of Toxaphene and Other Organochlorine Pesticides in North America As Determined by Air Measurements and Potential Source Contribution Function Analyses EUNHA HOH AND RONALD A. HITES* School of Public and Environmental Affairs, Indiana University, Bloomington, Indiana 47405

A previous study from our laboratory suggested that the Gulf of Mexico might be a potential source of toxaphene to the United States. To investigate this hypothesis, we measured gas-phase toxaphene concentrations at sampling sites ranging from northern Michigan to southern Louisiana; the samples were collected every 12 days during 2002-2003. We also measured other organochlorine pesticides in these samples. We identified major source regions of each pesticide group using the potential source contribution function model with the Clausius-Clapeyron equation defining the criterion levels. These results indicate that southern cotton farms are major sources of both toxaphene and p,p′-DDE to the northern United States. In fact, there is a very strong correlation of toxaphene and DDE atmospheric concentrations at all sites, further indicating a common source. On the other hand, the Gulf of Mexico is not a major source of toxaphene or DDE. DDE’s source region is similar to that of toxaphene but somewhat broader, reflecting DDT’s historically more diversified use. The level of endosulfan in the atmosphere at all of the sites was similar, and PSCF modeling indicated that its sources were all toward the east of the sampling sites.

Introduction Toxaphene was first used as an insecticide in the United States in 1947, but its peak consumption occurred in 1972 just after DDT was banned (1). The U.S. EPA severely restricted toxaphene’s use in 1982 due to its potential carcinogenicity and persistence. The limited use of existing stocks was allowed until 1986, and all uses were banned in 1990 (2). Historically, more than 85% of the toxaphene ever produced in the United States was used to control insects on cotton in the southern United States (3). In fact, the heavy application of toxaphene in the southern United States has made this a significant source region, which is responsible for the contamination of remote locations, such as the upper Great Lakes, by way of long-range atmospheric transport (4-7). Recently, James and Hites (8), using backward air trajectory analysis, showed that the airborne toxaphene found in Indiana and in the Great Lakes region originated in the southern United States. In addition to the cotton-growing regions of Arkansas, James and Hites suggested that the Gulf of Mexico also might be a potential source of toxaphene (8). * Corresponding author e-mail: [email protected]. 10.1021/es0499290 CCC: $27.50 Published on Web 06/30/2004

 2004 American Chemical Society

The objective of this study was to determine if the Gulf of Mexico was a significant source of toxaphene to the United States by the direct measurement of atmospheric toxaphene concentrations on the shore of the Gulf. To this end, we established a sampling site in Louisiana near the Gulf of Mexico coast, and we measured toxaphene at this site and at three other, previously established sampling sites in Arkansas, Indiana, and Michigan. Samples were taken at all four sites every 12 days from February 2002 to September 2003. To fully understand the behavior of toxaphene at each site, other organochlorine pesticides (p,p′-DDE, chlordanerelated compounds, and endosulfan-related compounds) were also measured over the same time period. To identify the sources of toxaphene and the other organochlorine pesticides coming to the sampling sites, the potential source contribution function (PSCF) method was used. The PSCF model is a probabilistic model based on calculated backward air trajectories, which has been primarily used to identify source regions of atmospheric particle-phase species, such as sulfate (9-12). Recently, Hsu et al. used this model to identify PCB sources in Chicago (13), and our laboratory used it to identify sources of PCBs and other persistent organic pollutants coming to the Great Lakes (14). In this study, we use this same approach and map potential source regions for toxaphene and other organochlorine pesticides in the United States.

Experimental Section Atmospheric Sample Collection. Air samples were collected at four sampling sites: (a) the Integrated Atmospheric Deposition Network (IADN) site located near Sleeping Bear Dunes National Lakeshore on the northeastern shore of Lake Michigan (44°48′47′′ N, 86°03′32′′ W), (b) Indiana University in Bloomington, IN (39°10′00′′ N, 86°31′17′′ W), (c) the University of Arkansas Southeast Research and Extension Center near Rohwer, AR (33°45′39′′ N, 91°16′32′′ W), and (d) the Louisiana Universities Marine Consortium in Cocodrie, LA (29°15′14′′ N, 90°39′4′′ W). The sampling site locations are shown in Figure 1. The Michigan sampler was located ∼1 km from the shore of Lake Michigan and about 15 km south of the town of Empire (population 500). The Indiana sampler was located on a porch adjacent to the School of Public and Environmental Affairs building, which was built in 1980, and is located on the northeast side of the city of Bloomington (population 60 000) on the Indiana University campus. The Arkansas sampler was located in the backyard of the Southeast Research and Extension Center’s office, which is a converted ranch-style house built in 1960, located on a fully functioning experimental agriculture farm ∼2 km north of the town of Rohwer (population 150). In Louisiana, the sampler was located on the roof of the two-story DeFelice Marine Center building built in 1986, which is located in Cocodrie, LA (population 25), approximately 135 km southwest of New Orleans. This building is located ∼17 km from the Gulf of Mexico, with Terrebonne Bay between the sampling site and the Gulf. The Louisiana site was chosen because toxaphene was not used in this region (3) and because this site would give air samples that are representative of air coming from the Gulf of Mexico (8). The other three sites were selected because they had previously been used by James and Hites (8) and by the IADN project (http://www.msc-smc.ec.gc.ca/iadn/ index_e.html). Specifically, the Arkansas site represents a region where toxaphene was heavily used, the Michigan site represents a remote area of the Great Lakes, and the Indiana VOL. 38, NO. 15, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4187

FIGURE 1. Map of the central United States indicating the sampling sites at Sleeping Bear Dunes, MI (MI); Bloomington, IN (IN); Rohwer, AR (AR); and Cocodrie, LA (LA). site is a middle point between the Great Lakes and the Arkansas site. The Arkansas and Louisiana samples were collected using Anderson high-volume samplers (Anderson Instrument Inc., Smyrna, GA; model PS-1). An older, Sierra-Misco version of the sampler was used at the Indiana and Michigan sites. All of these samplers draw air through a glass-fiber filter (Whatman, Clifton, NJ) to collect the particle-bound compounds and then through a polyurethane foam (PUF) adsorbent (Tisch Environmental, Inc., Village of Cleves, OH) to collect the gas-phase compounds. Sampling took place for 1 day every 12 days from February 2002 until September 2003. Sampling dates at all of the sites follow the IADN schedule, and sampling occurred from 9:00 a.m. to 9:00 a.m. Sample volumes for Arkansas and Louisiana were ∼400 m3 of air, and for Indiana and Michigan, the volumes were ∼1300 m3. The air volume for the southern samples (Arkansas and Louisiana) was increased to about 800 m3 from October 2002 by doubling the running time to 2 days. Site operators at each site installed the sampling media on the samplers, and after sampling, they shipped the samples to Indiana University for analysis. Only the gas-phase samples (the PUF adsorbents) were analyzed for toxaphene and the other organochlorine pesticides. Sample Preparation. Polyurethane foam adsorbents were precleaned by Soxhlet extraction for 24 h with 1:1 petroleum ether-acetone. The cleanup method was later changed. In this case, the PUF was washed with water, sonicated for 40 min, and then sequentially Soxhlet extracted for 24 h with each of the following solvents: methanol, acetone, hexane, dichloromethane, hexane, and 1:1 acetone-hexane. After sampling, the PUF samples were spiked with γ-13C10chlordane (Cambridge Isotope Laboratories, Andover, MA) and Soxhlet extracted for 24 h with 1:1 acetone-hexane. After extraction, the solvent was exchanged to hexane, reduced in 4188

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 15, 2004

volume to about 1 mL by rotary evaporation, and fractioned on 1% water-deactivated silica gel. Toxaphene and the other organochlorine pesticides were eluted with 3:2 hexanedichloromethane and then with 100% dichloromethane. These two fractions were combined and reduced in volume to 50 µL by N2 blow-down. The final extracts were spiked with a quantitation standard, 2,2′,3,4,4′,5,6,6′-octachlorobiphenyl (PCB-204) (AccuStandard Inc., New Haven, CT), prior to analysis. Instrumental Analysis. Toxaphene and the pesticides were quantitated by gas chromatographic mass spectrometry (Agilent 5973 mass spectrometer) operating in the electroncapture negative-ionization mode. The extracts were injected into an Agilent 6890 gas chromatograph fitted with a 60 m DB-5-MS column (250 µm i.d., 0.25 µm film thickness; J&W Scientific, Folsom, CA) in 2 µL volumes, using the splitless mode and helium as the carrier gas. The injection port was maintained at 285 °C. Each sample was run twice, once for toxaphene analysis and once for the other pesticides, using different temperature programs on the same instrument. The GC temperature program for toxaphene analysis began at 80 °C for 1 min; then it was ramped at 10 °C/min to 210 °C, ramped at 0.8 °C/min to 250 °C, and ramped at 10 °C/min to 310 °C, where it was held for 10 min. The total run time was 81 min. The GC temperature program for the organochlorine pesticides began at 40 °C for 1 min; the temperature then was ramped at 30 °C/min to 130 °C, at 3 °C/min to 241 °C, at 30 °C/min to 285 °C, where it was held for 20 min, and then at 30 °C/min to 300 °C, where it was held for 20 min. The total running time was 83 min. The GC to MS transfer line was heated to 280 °C, and the ion source of the mass spectrometer was held at 150 °C. The electron-capture negative-ionization GC/MS selected ion monitoring (SIM) method for toxaphene was developed by Swackhamer et al. and modified slightly for subsequent use (15, 16). The M- or (M - Cl)- ions of the hexa- through decachlorinated bornanes and camphenes were monitored in the SIM mode. Interference ions produced by chlordane and the 13C contributions from toxaphene fragment ions also were monitored. The background was subtracted, and the Agilent data analysis program generated an output file of peak areas and retention times of potential toxaphene peaks. A program, developed in our laboratory, selected the valid toxaphene peaks on the basis of chlorine isotope ratios and corrected for interfering compounds. Complete details are given elsewhere (16). Due to the complex nature of the toxaphene mixture, the relative response factor (RRF) is not linear over all concentration ranges. The calculated RRFs from each standard were plotted against the total peak area for that standard. A power function was fit to the data and used to calculate the RRF for each sample on the basis of its total toxaphene peak area. The sample was then analyzed using that RRF value. For the other organochlorine pesticide analyses, the most abundant ion in the molecular anion cluster was selected as a target ion, and the second most abundant ion in the molecular ion cluster was selected as a confirmation ion for each compound. The monitored ions (target ion and confirmation ion) are as follows: trans-chlordane (m/z 410 and 408), cis-heptachlor epoxide (388 and 390), oxychlordane (424 and 422), trans-nonachlor (444 and 442), endosulfan II (406 and 408), endosulfan sulfate (386 and 388), and p,p′DDE (318 and 316). The fragment ions 266 and 264 were chosen for cis-chlordane instead of the molecular ions 410 and 408, and the fragment ions 372 and 374 were chosen for endosulfan I instead of the molecular ions 406 and 408 because of the close retention times of these two compounds. Quantification of all pesticides was carried out using HP MS ChemStation software based on an external standard solution having known amounts of all the target compounds and

TABLE 1. Summary of the Concentration Data (pg/m3) and Clausius-Clapeyron Regression Results for Each Pesticide from Each Sampling Sitea pesticide

site

Cav ( std err

C288 ( std err

C median

range

n

∆H (kJ/mol)

r2

toxaphene

MI IN AR LA MI IN AR LA MI IN AR LA MI IN AR LA

23 ( 4 60 ( 10 1400 ( 200 61 ( 7 8.7 ( 1.5 6.4 ( 0.9 290 ( 40 3.6 ( 0.3 39 ( 7 210 ( 20 200 ( 20 59 ( 7 142 ( 45 260 ( 70 100 ( 20 100 ( 20

27 ( 3 39 ( 4 710 ( 77 36 ( 6 10 ( 2 5.1 ( 0.5 130 ( 16 ns 44 ( 6 190 ( 10 120 ( 10 27 ( 4 100 ( 18 95 ( 10 43 ( 4 23 ( 6

12 44 1100 45 4.5 5.6 200 3.4 18 200 160 50 37 79 60 52

0.90-110 3.0-360 69-4100 12-240 0.42-42 0.66-35 11-850 0.9-13 2.4-170 26-520 15-600 10-260 0.56-1200 2.7-2000 4.7-390 3.6-480

42 43 45 43 42 44 43 42 41 44 44 43 42 44 44 43

66 ( 6 85 ( 9 75 ( 9 31 ( 11 74 ( 8 51 ( 7 81 ( 10 ns 73 ( 7 57 ( 4 71 ( 7 52 ( 10 100 ( 9 120 ( 9 89 ( 8 78 ( 18

0.745 0.687 0.618 0.154 0.703 0.536 0.604 0.002 0.698 0.798 0.736 0.421 0.753 0.826 0.737 0.314

p,p′-DDE

chlordanes

endosulfans

a

See Figure 1. ns, no significant correlation.

surrogate and quantitation standards. The chlordane-related compound concentrations are given as the sum of the three most abundant components of technical chlordane (cis- and trans-chlordane and trans-nonachlor) and of the two chlordane transformation products (heptachlor epoxide and oxychlordane). The endosulfan-related compound concentrations are given as the sum of endosulfan I and II and their metabolite, endosulfan sulfate. Quality Assurance. All solvents were spectroscopic grade. Silica was pre-extracted using dichloromethane for 24 h, and the Na2SO4, glass wool, and disposable pipets were heated at 450 °C for a minimum of 6 h prior to use. Either a procedural blank or a spike recovery sample containing toxaphene (Hercules Co.) and a technical pesticide mixture (Dr. Ehrenstorfer) was run with every batch of eight samples. The recovery of toxaphene in the spike recovery samples was 102 ( 8% (n ) 23), and the recoveries of the other pesticides in the spike recovery samples were 88-98% with standard deviations of 9-19% (n ) 10). The recovery of the surrogate standard in all the samples was 108 ( 12% (n ) 219). The recoveries were high enough so that we did not correct the concentrations in the samples for the loss of the surrogate standard. In laboratory and field blanks, toxaphene was not detected. Only very small amounts of the other pesticides were detected in the blanks, so no blank corrections were performed. Air Trajectory Generation. The hybrid single-particle Lagrangian integrated trajectory (HYSPLIT) model was used to generate backward trajectories; this model is available on the National Oceanic and Atmospheric Administration (NOAA) Air Resource Laboratory website (www.arl.noaa.gov/ ready/hysplit4.html). Four day backward trajectories were generated at four different starting times during the 24 h sampling period; these starting times were at 6 h intervals when the sampling period was 24 h and at 12 h intervals when the sampling period was 48 h. In each case, the trajectories were calculated at three different starting altitudes: 10, 100, and 500 m above ground level. Thus, 12 backward trajectories were generated for each sample. Because each trajectory usually consisted of 96 hourly latitude-longitude coordinates, each chemical measurement was represented by ∼1152 such coordinates. About 500 trajectories were generated for each sampling site. The calculations used the Eta Data Assimilation System database of archived meteorological data. Potential Source Contribution Function. We implemented the PSCF model with a Microsoft visual basic program written in our laboratory (14). The Clausius-Clapeyron

equation (a linear regression between the natural logarithms of the partial pressures and the inverse of the atmospheric temperatures) for each pesticide group at each site was used as the criterion value. The mean value was used as the criterion value for p,p′-DDE at the Louisiana site because these atmospheric concentrations did not vary significantly with temperature. The residual was calculated from the criterion value for each sample; positive residuals correspond to “high” sampling days, and negative residuals correspond to “low” sampling days. The standard deviations of the residuals were also calculated for each pesticide group at each site. This PSCF program removed ambiguous samples, which were close to the criterion values (falling between (0.125 standard deviations from zero), and sorted all of the high and low hourly points from the HYSPLIT trajectories onto a geographical 1° × 1° cell. The program calculated the PSCF value in each grid cell by

PSCF )

nhigh nhigh + nlow

(1)

where nhigh is the number of unambiguously high concentration hourly points in a given cell and nlow is the number of unambiguously low concentration hourly points in a given cell. All of these values were plotted vs the cell locations using mapping software (ArcGIS 8.2). Only the cells including at least 25 points or more were plotted.

Results and Discussion Atmospheric Organochlorine Pesticide Concentrations. The arithmetic average concentrations of the various analytes (with their standard errors) in the atmospheric samples collected at each sampling site are given in Table 1. The temperature-corrected concentrations (C288), the median concentrations, and the range of the concentrations are also summarized in Table 1. The temperature-corrected concentrations were calculated from the regression equation between the natural logarithm of the compound’s partial pressure and the inverse of the average temperature during sampling:

P288 ) P exp

[∆HR(T1 - 2881 )]

(2)

where P is the partial pressure of the compound (atm), ∆H is the phase-transition energy, R is the gas constant, and T is the atmospheric temperature (K). In reality, ∆H is a composite value, representing the energy needed to move VOL. 38, NO. 15, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4189

FIGURE 2. Temperature dependence of organochlorinated pesticide partial pressure at all four sampling sites: (A) toxaphene (b, AR; O, MI, IN, and LA); (B) p,p′-DDE (b, AR; O, MI, IN, and LA); (C) chlordanes (b, AR and IN; O, MI and LA); (D) endosulfans (b, all sites). 1 mol of each pesticide group from soil, vegetation, particles, or water into the gas phase. This equation was applied for each pesticide group at each sampling site in this study, and the ∆H values and the linear regression coefficient values (r2) are given in Table 1. The correlation between the natural logarithms of the partial pressures of all pesticide groups and the reciprocal temperature is significant at the 99% confidence level for all four sites, except for p,p′-DDE at the Louisiana site. The average concentration of toxaphene is by far the highest at the Arkansas site. The arithmetic average concentration of toxaphene at the Arkansas site was 20-60 times higher than that at the other sites. It is clear that this region is a major source region of toxaphene to the atmosphere. Surprisingly, the average toxaphene concentration at the Louisiana site was about the same as at the Indiana site. This result suggests that the Gulf of Mexico is not a major toxaphene source. The arithmetic average and the median toxaphene concentrations at the Michigan site were lower than at the Indiana and Louisiana sites, but the temperaturecorrected values were about the same. In general, the toxaphene concentrations are similar at the Michigan, Indiana, and Louisiana sites. The high linearity of the combined Clausius-Clapeyron plot for these three sites (see Figure 2A) supports this conclusion. In the previous study, James and Hites (8) reported the temperature-corrected concentration of toxaphene at the same Michigan, Indiana, and Arkansas sites. Our concentrations at these sites are about a factor of 2 higher than these previously reported values for both the Michigan and Indiana sites and ∼25% lower at the Arkansas site. The lower concentrations at the Arkansas site may be the result of the two years that had passed between the two sets of measurements. A regression by James and Hites (8) suggested that toxaphene concentrations would decrease by ∼35% between 2000 and 2002, and this is about what we observed. The reason for the increased concentrations at the Michigan and Indiana sites is not clear, but we note that these are very low toxaphene concentrations and that measurement error may be significant at these levels. Taken as a whole, the toxaphene concentrations we measured are similar to concentrations reported for other locations in North America (6, 8, 17-20). p,p′-DDE is the major environmental degradation product of DDT, and DDE is more volatile and environmentally persistent than its parent compound. The arithmetic average concentration of DDE at the Arkansas site was 45-70 times higher than that at the other sampling sites (see Table 1). The temperature-corrected and median concentrations of 4190

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 15, 2004

FIGURE 3. Relationship of p,p′-DDE and toxaphene concentrations at all sampling sites. The dashed lines represent 99% confidence intervals. p,p′-DDE at the Arkansas site were also much higher than at the other sites. The DDE concentrations were similar at the Michigan and Indiana sites, but the DDE concentrations at the Louisiana site were slightly lower than at the Michigan and Indiana sites. Interestingly, atmospheric temperature was not a significant influence on DDE atmospheric concentrations at the Louisiana site. Because the DDE concentration ranges for the Michigan, Indiana, and Louisiana sites are similar, the concentrations for all samples from the three sites were plotted together and the Clausius-Clapeyron equation was fitted to these data (see Figure 2B). The correlation is significant at the 99% confidence level and suggests that the DDE atmospheric concentrations at the three sites are similar. The DDE concentrations we measured at the Michigan, Indiana, and Louisiana sites are similar to values previously measured in air sampled from North America (17, 18, 21, 22). The Arkansas value is very much higher and suggests that this region is a source of DDE to the atmosphere, as it was for toxaphene. That this location should be a major source region 30 years after the banning of DDT in the United States is a testament to the persistence of this compound. The concentrations of total chlordane were highest at the Indiana and the Arkansas sites (see Table 1), and these may be source regions to the rest of North America. The concentrations of chlordane at the Louisiana and Michigan sites were similar to one another and about a factor of 4 lower than at the Indiana and Arkansas sites. On the basis of these observations, the Clausius-Clapeyron equation was applied for the samples from the Arkansans and Indiana sites together and for the samples from the Louisiana and Michigan sites together. This approach gave high linearity for each regression; see Figure 2C. It is surprising that the chlordane concentrations at the Indiana site (not a particularly active agricultural region) were similar to the concentrations at the Arkansas site (an intensely agricultural site). Perhaps chlordane had been used as a termiticide around buildings near the Indiana site. The concentrations (and their ranges) we measured here are similar to total chlordane concentrations measured throughout North America (1719, 21, 22). The arithmetic average concentration of total endosulfans was highest at the Indiana site and about the same at the other three sites. The temperature-corrected endosulfan concentrations at the Michigan and Indiana sites were similar and about 3 times higher than those at the other two sites, which were similar to each other. The endosulfan concentrations in the four samples taken during June and July at the Michigan site were 690-1200 pg/m3, which were much higher

FIGURE 4. Individual PSCF maps for toxaphene: (A) Michigan; (B) Indiana; (C) Arkansas; (D) Louisiana. Dark red represents PSCF values from 0.91 to 1.0, with shades of pink ranging down to 0.61 by tenths, and white represents from 0.41 to 0.60. Dark blue represents 0-0.10, with shades of blue ranging up to 0.40 by tenths. The green star in each map represents the sampling site.

FIGURE 5. Individual PSCF maps for p,p′-DDE: (A) Michigan; (B) Indiana; (C) Arkansas; (D) Louisiana. See Figure 4 for color codes. The green star in each map represents the sampling site. VOL. 38, NO. 15, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4191

FIGURE 6. (A) PSCF map for endosulfan at all of the sampling sites used in this paper. (B) PSCF map for endosulfan from IADN samples (10). See Figure 4 for color codes. The green stars in the maps represent sampling sites. than for the rest of the samples at that site. Likewise, the endosulfan concentrations in the five samples collected during July and August at the Indiana site were 1020-2000 pg/m3, which were much higher than for the rest of the samples at that site. Given that endosulfan is still used (unlike the other compounds measured in this study), it is possible that endosulfan was being applied during the June to August period to crops grown in the vicinity of the Michigan and Indiana sites. Excluding these few high samples, the concentration ranges for all the sites were similar, and the Clausius-Clapeyron equation was fitted to the data from all of the sites combined (see Figure 2D). The similar endosulfan concentrations among the sites and the high concentrations at the Michigan and Indiana sites during the summer imply that endosulfan’s source may be direct agricultural application around the sites. Recently, Buehler et al. showed that the agricultural application of endosulfan in June to August was a major factor contributing to the variability of its atmospheric concentrations near the Great Lakes (23). These dates agree with the times when we observed high endosulfan concentrations. Interestingly, ∆H of toxaphene at the Louisiana site is significantly lower than that at the other sites, and ∆H of p,p′-DDE at this site is not significant. Jantunen et al. reported 4192

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 15, 2004

a similar result for p,p′-DDE in the air of northwestern Alabama during 1996-1997 (18). The higher ∆H values of endosulfan at the Michigan and Indiana sites may indicate that endosulfan is used near the two northern sites during the summer. The overall values of the phase-transfer energies for each pesticide among the sites are within the ranges measured in North America by other studies (5, 6, 8, 17-20, 23-26). Comparing the linear regression coefficients (r2) for the Clausius-Clapeyron regression for each pesticide group among the sites, we note that the r2 values from the Louisiana samples are always much lower than for the other sites. One reason for this difference may be the shorter atmospheric temperature range observed in Louisiana. In this case, the temperature range was 287-302 K (with two exceptions), while the temperature ranges were 262-300, 268-298, and 273-303 K at the Michigan, Indiana, and Arkansas sites, respectively. This small temperature range may be a result of the regional characteristics of the Louisiana site, which was close to the Gulf of Mexico. At this location, the air from the sea lowers the temperature in the summer and increases it in the winter. The concentration patterns are similar between p,p′-DDE and toxaphene at the sampling sites; for example, the

FIGURE 7. Total endosulfan usage in 1997 for the eastern United States (data from http://www.ncfap.org/). atmospheric concentrations of both pesticides were highest at the Arkansas site. Figure 3 shows the relationship between the concentrations of p,p′-DDE and the concentrations of toxaphene in all of our samples at all sites on a log-log scale. This graph shows a strong correlation between these pesticide concentrations and suggests a common source for both. This result is perhaps not surprising given that both of these pesticides were heavily used on cotton. For example, over 80% of DDT’s use in the United States was on cotton during the period 1970-1972, and a 2:1 mixture of toxaphene and DDT was used on cotton in the mid-1970s (27). Eventually, toxaphene was the recommended replacement for DDT after it was banned in 1972 in the United States. Sources of Each Pesticide Group Using the PSCF Model. Maps of potential toxaphene source regions were produced for each of the sites and are shown in Figure 4. For Michigan and Indiana (see Figure 4A,B), the maps indicate that toxaphene is coming to the sampling site from the south, presumably from the cotton-growing regions centered in Arkansas. This result agrees with the previous study by James and Hites, who determined that the toxaphene detected in the air in Indiana and in the Great Lakes region originated in the southern United States by using a simple nonparametric directional indicator (8). As discussed above, the atmospheric concentration of toxaphene is high at the Arkansas site, which indicates that this is a toxaphene source region. The PSCF map for this site (see Figure 4C), however, indicates that there may be toxaphene source regions to the southwest of this location in Louisiana, Texas, and Oklahoma and that the Gulf of Mexico itself may be a toxaphene source to the Arkansas site. This result agrees with the previous study by James and Hites (8). On the other hand, the PSCF map for the Louisiana site (see Figure 4D) indicates that the Gulf of Mexico is not a toxaphene source. In this case, the PSCF model indicates that the toxaphene sources to the Louisiana site are north of the site in states ranging from Texas to Georgia. It seems likely that there are significant toxaphene sources in northern

Louisiana that impact the Arkansas site when the air is coming from the south but that the Gulf of Mexico is not a source. We constructed the same type of PSCF maps for p,p′DDE; see Figure 5. These maps were similar to the PSCF maps for atmospheric toxaphene concentrations with a couple of exceptions. The map for the Michigan site (see Figure 5A) shows a DDT source region in southern Michigan and Ontario, a result which is compatible with the use of DDT in fruit orchards in Michigan and Ontario (28, 29). The Arkansas site is a source itself, but the PSCF map (see Figure 5C) indicates that the Gulf of Mexico could also be a source region for this location. However, like toxaphene, the PSCF map for DDE to the Louisiana site (see Figure 5D) shows that the Gulf of Mexico is not a DDE source. Interestingly, Florida seems to be a potential source region of DDE to the Louisiana site but not of toxaphene; compare Figures 4D and 5D. This observation is compatible with the historical use of DDT for agriculture and insect control in Florida (30). In addition, the Tower Chemical Co. spilled a large amount of dicofol to Lake Apopka in 1980 (31). At that time, dicofol was contaminated with ∼15% DDT and its metabolites (32). Given that dicofol was heavily used on cotton and citrus crops in Florida, it is possible that this is an additional source of DDE localized in Florida. Given the similarity of the endosulfan concentrations among the sites, all four sampling sites were included on one PSCF map. High and low residuals were calculated separately from each Clausius-Clapeyron regression at each site, combined, and processed through the visual basic PSCF model. Due to a higher number of trajectories (153 samples × 4 different times × 3 heights), 0.5° × 0.5° was chosen as a cell size for the PSCF map in Figure 6A. This map shows a strong directional trend with a large potential endosulfan source region running from Kentucky through Tennessee, Alabama, and into Florida. Conversely, there seems to be no source region to the west of the sampling sites. Comparing this map with the endosulfan PSCF map generated by Hafner and Hites using IADN data (see Figure 6B), we note that the VOL. 38, NO. 15, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4193

coverage of our study extends further to the south, while their map covers a wider area of the northern United States. Taken together, these two maps indicate endosulfan sources in Kentucky through Florida and in New York. On the basis of a database of endosulfan use recorded by states from the National Center for Food and Agricultural Pesticides, we generated a map showing endosulfan usage in 1997 in the states covered by our PSCF maps (see Figure 7). The heaviest user states of endosulfan were Kentucky, Georgia, and Florida; the next highest endosulfan user states were Michigan and New York. These usage areas matched well with the potential source regions in the PSCF maps (see Figure 6). The exception is Alabama, which was not a heavy user of endosulfan in 1997, despite the suggestion of our PSCF map. In this case, endosulfan may be simply passing through Alabama in air transported from adjacent states, or endosulfan could have become more heavily used in Alabama after 1997. The chlordane PSCF maps did not show any significant trends, so these maps are not included here.

Acknowledgments We thank Amy Wilson-Finelli and Dr. Rodney Powell at Louisiana Universities Marine Consortium, Randy Spurlock and Larry Earnest at the University of Arkansas Southeast Research and Extension Center, and Tom and Alice Van Zoeren at Sleeping Bear Dunes National Lakeshore for acting as sampling site operators. We also thank NOAA’s Air Resources Laboratory for providing the HYSPLIT transport model and William Hafner for help with the PSCF analysis. This project was supported, in part, by the Great Lakes National Program Office of the U.S. Environmental Protection Agency (Grant GL995656, Melissa Hulting, Project Officer).

Supporting Information Available A table giving the gas-phase atmospheric toxaphene, total chlordane, total endosulfan, and p,p′-DDE concntrations and average air temperature in the sampling sites. This material is available free of charge via the Internet at http:// pubs.acs.org.

Literature Cited (1) (2) (3) (4)

Saleh, M. A. Rev. Environ. Contam. Toxicol. 1991, 118, 1-85. Fed. Regist. 1982, 47, 53784. Li, Y. F. J. Geophys. Res. 2001, 106, 17919-17927. Rice, C. P.; Samson, P. J.; Noguchi, G. E. Environ. Sci. Technol. 1986, 20, 1109-1116. (5) Glassmeyer, S. T.; Brice, K. A.; Hites, R. A. J. Great Lakes Res. 1999, 25, 492-499. (6) James, R. R.; McDonald, J. G.; Symonik, D. M.; Swackhamer, D. L.; Hites, R. A. Environ. Sci. Technol. 2001, 35, 3653-3660. (7) Rice, C. P.; Evans, M. S. Adv. Environ. Sci. Technol. 1984, 14, 163-194.

4194

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 15, 2004

(8) James, R. R.; Hites, R. A. Environ. Sci. Technol. 2002, 36, 34743481. (9) Ashbaugh, L. L.; Malm, W. C.; Sadeh, W. Z. Atmos. Environ. 1985, 19, 1263-1270. (10) Cheng, M.; Hopke, P. K.; Barrie, L.; Rippe, A.; Olson, M.; Landsberger, S. Environ. Sci. Technol. 1993, 27, 2063-2071. (11) Lupu, A.; Maenhaut, W. Atmos. Environ. 2002, 36, 5607-5618. (12) Polissar, A. V.; Hopke, P. K. Environ. Sci. Technol. 2001, 35, 4214-4226. (13) Hsu, Y.-K.; Holsen, T. M.; Hopke, P. K. Environ. Sci. Technol. 2003, 37, 681-690. (14) Hafner, W. D.; Hites, R. A. Environ. Sci. Technol. 2003, 37, 37643773. (15) Swackhamer, D. L.; Charles, M. J.; Hites, R. A. Anal. Chem. 1987, 59, 913-917. (16) Glassmeyer, S. T.; Shanks, K. E.; Hites, R. A. Anal. Chem. 1999, 71, 1448-1453. (17) Alegria, H. A.; Bidleman, T. F.; Shaw, T. J. Environ. Sci. Technol. 2000, 34, 1953-1958. (18) Jantunen, L. M.; Bidleman, T. F.; Harner, T.; Parkhurst, W. J. Environ. Sci. Technol. 2000, 34, 5097-5105. (19) Bidleman, T. F.; Alegria, H.; Ngabe, B.; Green, C. Atmos. Environ. 1998, 32, 1849-1856. (20) Shoeib, M.; Brice, K. A.; Hoff, R. M. Chemosphere 1999, 39, 849871. (21) Buehler, S. S.; Basu, I.; Hites, R. A. Environ. Sci. Technol. 2001, 35, 2417-2422. (22) Park, J.; Wade, T. L.; Sweet, S. Atmos. Environ. 2001, 35, 33153324. (23) Buehler, S. S.; Basu, I.; Hites, R. A. Environ. Sci. Technol. 2004, 38, 414-422. (24) Cortes, D. R.; Basu, I.; Sweet, C. W.; Brice, K. A.; Hoff, R. M.; Hites, R. A. Environ. Sci. Technol. 1998, 32, 1920-1927. (25) Burgoyne, T. W.; Hites, R. A. Environ. Sci. Technol. 1993, 27, 910-914. (26) Hoff, R. M.; Muir, D. C. G.; Grift, N. P. Environ. Sci. Technol. 1992, 26, 276-283. (27) U.S. Environmental Protection Agency. DDT Regulatory History: A Brief Survey (to 1975); http://www.epa.gov/cgi-bin/ epaprintonly.cgi. (28) Harris, M. L.; Wilson, L. K.; Elliott, J. E.; Bishop, C. A.; Tomlin, A. D.; Henning, K. V. Arch. Environ. Contam. Toxicol. 2000, 39, 205-220. (29) Meijer, S. N.; Shoeib, M.; Jantunen, L. M.; Jones, K. C.; Harner, T. Environ. Sci. Technol. 2003, 37, 1292-1299. (30) Majewski, S. N.; Capel, P. D. Pesticides in the Atmosphere: Distribution, Trends, and Governing Factors; Ann Arbor Press: Chelsea, MI, 1995. (31) Clark, D. R. Dicofol as an Environmental Contaminant; U.S. Fish and Wildlife Technical Report 29; U.S. Fish and Wildlife Service: Washington, DC, 1990. (32) Jarman, W. M.; Springer, A. M.; Walker, W. A., II; Hunt, W. G. Environ. Toxicol. Chem. 1986, 5, 1319.

Received for review January 13, 2004. Revised manuscript received May 11, 2004. Accepted May 14, 2004. ES0499290